Deconvolution of complex spectra outsourced to algorithm

Skip to Navigation


  • Published: Oct 15, 2016
  • Author: Ryan De Vooght-Johnson
  • Channels: Laboratory Informatics / Chemometrics & Informatics
thumbnail image: Deconvolution of complex spectra outsourced to algorithm

Needle in a haystack

A new algorithm for spectral deconvolution and xenobiotic identification aids forensic toxicology investigations, uncovering more xenobiotics in blood samples than humans alone.

Blood is notoriously difficult to analyse. For one, the 22 most abundant proteins account for ninety-nine percent of all proteins in circulation. These proteins, much like the unwanted stones that clogged the sieves of gold-rush miners, will mask even the faintest glimmering speck at the bottom of your pan.

Further still, if you want to detect xenobiotics at their trace therapeutic levels—and perhaps even at sub-therapeutic levels—you’d better hope that your test is sensitive enough. Surveying the blood for your xenobiotic really is like searching for a needle in a haystack.

Whilst clinical toxicologists can surmount this challenge by shunning blood for urine (if viable), there are no alternatives for forensic toxicologists. Urine, after all, may not be available on tap at the site of a road collision. Blood invariably is.

Expert vs algorithm

Having received the spun-down serum—that is whole blood minus the red and white blood cells and clotting factors—it must be screened for the xenobiotic(s) of interest before a definitive quantitation is taken. Naturally, immunoassays or some form of basic detector—perhaps a diode array—are used in combination with LC.

Recently, however, chromatography paired up with a mass spectrometer has fit the billing. Fortunately for GC-MS-based analyses, there exists a large reference pharmacopoeia to which mass spectra can be compared to and xenobiotics identified: The Maurer/Pfleger/Weber MPW_2011.

If, however, you wished to interpret the reams of highly convoluted mass spectra generated in pursuit of your xenobiotic (remember all those proteins?), you’d need over 10,000 hours experience and a lot of time or you must use a computer-based algorithm.

‘Low dose drugs overlaid by matrix compounds are difficult to detect,’ explains Marcel Grapp, lead researcher of the study from the Georg-August and Saarland Universities of Germany. ‘To overcome this problem,’ he adds, ‘deconvolution algorithms are needed that extract pure compound peaks from complex matrix in the total ion chromatogram.’


Writing in the journal Drug Testing and Analysis, Grapp took 150 serum samples submitted by the local police force and, after a quick neutral and basic liquid–liquid extraction, submitted these for non-targeted screening by GC-MS.

He then handed over the spectral deconvolution and xenobiotic identification to a freeware algorithm—AMDIS for short—and compared its findings with those derived from semi-automated and manual labour. All three were directed to interpret their mass spectra in comparison to those contained within the Maurer/Pfleger/Weber MPW_2011 library.

The results were staggering. The AMDIS algorithm identified xenobiotics that would have otherwise been overlooked by a human interpreter in 25 of all samples tested (17%). What’s more, these extra IDs would be pertinent to most police investigations: antidepressants—citalopram, mirtazapine, venlafaxine—and the antipsychotic quetiapine. The levels of these drugs spanned from the trace sub-therapeutic to toxic overdose levels.

In addition to these, the algorithm detected metabolites buried deep within the complex spectra, which it were then used to identify the un-metabolised parents: tramadol (the hydroxyl metabolite was detected) and quetiapine (O-dealkyl-carboxy quetiapine detected). These, too, were overlooked by the human expert.

‘With assistance of AMDIS,’ Grapp concludes, ‘the number of forensic relevant target hits increased by 17%...due to better identification of low-abundant peaks in the TIC [total ion chromatogram] that could be easily overlooked by manual data examination.’

Nevertheless, we have not yet arrived at fully automated data processing: ‘AMDIS-based data evaluation still requires visual inspection and verification of the data by an experienced toxicologist.’

Related Links

Drug Testing and Analysis, 2016, 8, 816–825.. Grapp et al. Systematic forensic toxicological analysis by GC-MS in serum using automated mass spectral deconvolution and identification system.

Article by Ryan De Vooght-Johnson

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.

Follow us on Twitter!

Social Links

Share This Links

Bookmark and Share


Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Most Viewed

Copyright Information

Interested in separation science? Visit our sister site

Copyright © 2018 John Wiley & Sons, Inc. All Rights Reserved